The Nifty traded back to the previous day's close on 55 of the 97 sessions I checked, between 15 May and 1 October 2026. That headline rate of 57% hides the useful part: the size of the gap decided most of it. Gaps under 25 points filled on 23 of 27 days. Gaps of 100 points or more filled on 8 of 29.
Today was a near miss that shows how fine the line is. The Nifty opened 47.50 points higher at 22,603.25, and its low, 22,561.60, stopped 5.85 points short of Monday's close of 22,555.75 (today's post-market report). Everything below is a count of past days, not a forecast.
The question and the data
The question. When the Nifty opens away from the previous close, how often does it trade back to that close during the same session, and does the answer change with the direction and size of the gap?
The data. Daily open, high, low, close and previous close for the Nifty 50 index from my own market log. The log starts on 14 May 2026, and that first day has no previous close, so the study covers 97 sessions, from 15 May to 1 October 2026. I checked the last close in the log, 22,421.95 on 1 October, against Zerodha Kite and it matches.
Two things about the sample matter before any numbers:
- 97 days is small. One day moving between groups shifts a group's percentage by several points, and some groups below have only 6 or 10 days.
- The period leaned down. The Nifty closed at 23,713.75 on 14 May and 22,421.95 on 1 October, a fall of 1,291.80 points (5.45%). Not a crash, but not a neutral stretch either.
How I counted
The gap is the open minus the previous close. Every day in the sample had one: 52 opened higher and 45 opened lower, and none opened exactly at the previous close. The smallest gap was 0.55 points and the largest 353.10 points.
A gap is filled if the day's low reached the previous close after a gap up, or the day's high reached it after a gap down. Touching counts. I didn't ask the index to close there or stay there.
I used daily highs and lows, so this count can't tell me when the fill happened, only whether it did. A fill at 9:16 and a fill at 3:29 count the same.
The open itself comes from NSE's pre-open auction, which I covered in the pre-open session guide. That matters because the gap is set before continuous trading starts, often in line with where GIFT Nifty was trading overnight.
Results: just over half of gaps filled
| Gap direction | Days | Filled | Share |
|---|---|---|---|
| Up | 52 | 32 | 62% |
| Down | 45 | 23 | 51% |
| All | 97 | 55 | 57% |
Gap-up days filled more often than gap-down days. Part of that is size: the average gap up was 74.6 points and the average gap down 83.9 points, and as the next table shows, bigger gaps filled less often. In a period when the index fell overall, gap-downs were also slightly larger and slightly less likely to retrace.
Gap size decided most of it
| Gap size | Days | Filled | Share |
|---|---|---|---|
| Under 25 points | 27 | 23 | 85% |
| 25 to 50 points | 19 | 14 | 74% |
| 50 to 100 points | 22 | 10 | 45% |
| 100 points or more | 29 | 8 | 28% |
This is the clearest pattern in the data. A small gap is a small distance, and the Nifty covers a few dozen points in most sessions: in my daily range study the median day's high-to-low range was about 170 points. A 20-point gap is inside the noise of an ordinary morning. A 150-point gap needs the index to give back a large share of a full day's movement.
Split by direction, the same shape shows up on both sides:
| Gap size | Gap up: filled / days | Gap down: filled / days |
|---|---|---|
| Under 25 points | 12 / 14 | 11 / 13 |
| 25 to 50 points | 10 / 13 | 4 / 6 |
| 50 to 100 points | 8 / 12 | 2 / 10 |
| 100 points or more | 2 / 13 | 6 / 16 |
Two cells stand out, and I hold both loosely because the groups are small. Gap-downs of 50 to 100 points filled on only 2 of 10 days, against 8 of 12 for gap-ups of that size. And for gaps of 100 points or more it flipped: gap-downs filled 6 of 16 times, gap-ups only 2 of 13. With groups this size, three or four days in either direction would change the story, so I wouldn't read a rule into either.
Today's gap of 47.50 points sat in the 25 to 50 group, where 14 of 19 days filled. It didn't. That's what a 74% rate looks like from the inside: one day in four, it doesn't happen.
Filling the gap is not the same as reversing
A fill only means the index touched the previous close. It doesn't mean the day closed on the other side of it. I counted that separately:
| Gap up (52 days) | Gap down (45 days) | |
|---|---|---|
| Gap filled during the day | 32 (62%) | 23 (51%) |
| Closed on the other side of the previous close | 17 (33%) | 10 (22%) |
| Closed beyond the open, in the gap's direction | 21 (40%) | 23 (51%) |
So on gap-up days, the index came back to the previous close on 32 days but closed below it on only 17. On gap-down days the gap filled 23 times and the index closed above the previous close 10 times. Many fills were a touch, not a turn.
The last row is the other side of the same coin. On 21 gap-up days the Nifty closed above its opening price, and on 23 gap-down days it closed below its open. The open was rarely the end of the move in either direction, and rarely a reliable start of one either.
This fits what I found in the opening range study: the early part of the session tests both sides often, and the first direction is a weak guide to the close.
Expiry days
25 sessions in the sample were weekly or monthly expiry days. The gap filled on 17 of them (68%), compared with 38 of the other 72 days (53%). That is consistent with the larger intraday swings I described in what changes on expiry days, but 25 days is too few to treat the difference as settled.
Limitations
- Small, one-directional sample. 97 sessions over four and a half months, in a market that fell 5.45%. A rising stretch could give a different split between gap-ups and gap-downs.
- Touch, not hold. A single tick at the previous close counts as a fill. A stricter rule, such as a 15-minute candle closing beyond the previous close, would count fewer fills.
- No timing. Daily bars can't say whether a gap filled in the first minute or the last. I haven't measured that yet.
- Fixed size buckets. I chose 25, 50 and 100 points before running the count and didn't try other cut-offs. Points also mean slightly different things as the index level changes; at these levels the difference is small.
- Index, not a tradeable price. These are Nifty 50 spot values. Futures and options carry premium, spreads and costs that this count ignores.
What I take from it
"Gaps get filled" is half true. In this sample it described small gaps well (85% under 25 points) and big gaps badly (28% at 100 points or more). The direction of the gap mattered less than its size, and a fill was usually a visit to the previous close rather than a reversal through it.
If I use gap data anywhere, it will be as context for the open, alongside where GIFT Nifty traded overnight, not as a signal on its own. Next, I want to add the time of the fill from minute bars, and rerun the counts once the log includes a rising stretch.
Sources
- Nifty 50 daily open, high, low, close and previous close: my own market log, 14 May to 1 October 2026 (aggregate counts only)
- Cross-check of the 1 October close, and the 6 October example: Zerodha Kite market data
This post describes an analysis of the sessions I analysed, for information only. It is not investment advice or a recommendation to trade. Past patterns say nothing certain about the future. I am not registered with SEBI as an investment adviser or research analyst.